Release notes

v4.0.0 (unreleased)

Breaking Changes

• The underlying data allowing self attration and loading (SAL) correction to tidal forcing has been updated for accuracy, and was derived from the TPXO10v2a datasets. This is not a functional breaking change, but will change (improve) output from TidalForcing based on TPXO data compared to previos versions of ROMS-Tools. See (#617)

• ERA5 data correction in SurfaceForcing now includes longwave radiation, and the shortwave radiation correction multipliers are updated (#614). This is a breaking change only in the sense that input radiation fields generated by ROMS-Tools using ERA5 forcing data will be slightly different (but more correct) compared to previous ROMS-Tools versions. Correction factors for previous versions of ROMS-Tools remain functional and available.

• pco2 is no longer read in during SurfaceForcing type='bgc' (#608).

• BoundaryForcing’s apply_2d_horizontal_fill argument is deprecated in favor of the new prefill argument (apply_2d_horizontal_fill=True → prefill="2d_lateral_fill", False → prefill=None); it still works but emits a DeprecationWarning. Note that prefill="2d_lateral_fill" now regrids with the default "auto" engine (xESMF when installed), so its output differs slightly from the pre-v4 AMG+scipy path; pass regrid_method="scipy" to reproduce the legacy output byte-for-byte. (#616).

• SurfaceForcing now regrids horizontally with the same NaN-aware approach as BoundaryForcing by default (all types — physics, bgc, restoring): masked xESMF bilinear regridding plus inverse-distance-weighted extrapolation (nearest-neighbor pre-fill + scipy linear interpolation fallback when xESMF is unavailable), replacing the AMG Poisson 2D fill + scipy interpolation. Output shifts slightly for users on the defaults; pass prefill="2d_lateral_fill", regrid_method="scipy" to reproduce the legacy output. On the default extrapolating path, a ROMS grid that extends beyond the source data coverage now raises a clear coverage error rather than being silently extrapolated. (#629)

• from_file is now a private function to Grid. Files are now loaded as Grid(filename='grid.nc') (#573)

• The ChildGrid has been removed. Both a child and parent grid are created using Grid, and the functions align_grids and make_edata are called to adjust bathymetry and do the mapping (#573).

• Default river/CDR tracer concentrations are now read from river_tracer_defaults.nc and differ from the previous hardcoded values (e.g. DIC, NO3, ALK, temperature) (#615)

• CDR metrics on ds_cdr rename efficiency fields: cdr_efficiency to cdr_efficiency_from_flux, and cdr_efficiency_from_delta_diff to cdr_efficiency_from_DIC_difference. Saved datasets, ensemble inputs, or downstream code using the old names need updating. (#591)

• CDR ensemble statistics on Ensemble.ds rename efficiency fields: ensemble_mean to ensemble_efficiency_mean and ensemble_std to ensemble_efficiency_std (#591).

• Drop support for Python 3.11.

New Features

• SurfaceForcing was refactored to be more dask-lazy, such that data loading/streaming is minimized until save is called on the class, and thereby supports longer periods between start/end dates.

• SurfaceForcing with type='bgc' has new data source option of ‘MBL_co2’ for time-varying co2; climatology no longer acceptd for xco2. (#608)

• Density-space vertical interpolation for BGC tracers in InitialConditions and BoundaryForcing via the new bgc_interpolation_method="density" option. The source density coordinate is built from the BGC dataset’s own temperature/salinity (the unified dataset provides temp_WOA/salt_WOA). The target density uses the model’s temperature/salinity: for InitialConditions this is the physics already in the same object, while for BoundaryForcing the physics BoundaryForcing must be passed as physics_forcing=. When a BGC source lacks temperature/salinity (e.g. CESM), interpolation falls back to depth space. A density BoundaryForcing’s physics_forcing companion is now embedded in and restored from its YAML, so to_yaml/from_yaml round-trips preserve density-space interpolation instead of silently falling back to depth (#620).

• BGC tracer vertical interpolation is now selected via a bgc_interpolation_method string parameter on InitialConditions and BoundaryForcing, accepting "depth" (default), "density", or "density_mld". The new "density_mld" method identifies the mixed layer depth (MLD) in the source and target density fields (potential density exceeding the surface value by 0.03 kg/m³, the xroms mld convention); the source mixed layer is scaled so its MLD matches the target’s, and below the MLD the tracer is interpolated 1:1 in depth. This keeps the source and target mixed layers aligned while preserving the absolute depth of sub-mixed-layer features, and avoids the surface degeneracy of pure density-space interpolation. For BoundaryForcing, the physics temperature/salinity that define the target density coordinate are sampled at the nearest time to each BGC record (sufficient given the climatological output cadence and ROMS’ linear-in-time boundary interpolation), which avoids loading the full physics time series into memory. (#626)

• make_edata changed to make_nesting_info

• to_yaml and from_yaml were adjusted to handle child grids after they’ve been modified (#573)

• Nesting now supports optional baroclinic pressure fluxes via metadata (#568)

• Include time records strictly outside start/end bounds for SurfaceForcing, BoundaryForcing (#547)

• Nesting operations now infer open boundaries from the mask by default and supports optional BGC boundary outputs via metadata (#550)

• Memory savings and speedup for vertical regridding in InitialConditions and BoundaryForcing (#528)

• Potential memory savings and speedup for reading of optimally-chunked netcdf source data files in InitialConditions, BoundaryForcing, and SurfaceForcing (#572)

• Option to automatically close narrow 1-cell water channels during mask generation via Grid.update_mask(close_narrow_channels=True) or Grid(close_narrow_channels=True); integrated into the standard mask workflow

• BoundaryForcing now regrids horizontally with a NaN-aware approach by default, removing the need for the slow and memory-hungry 2D Poisson horizontal fill on large multi-year domains. When xESMF is available, masked bilinear regridding (weights renormalized over valid ocean cells) plus inverse-distance-weighted extrapolation (configurable via the extrap_method / extrap_kwargs arguments) produces NaN-free boundaries directly. When xESMF is unavailable (e.g. Windows/pip installs), a nearest-neighbor pre-fill of the source followed by scipy linear interpolation provides an equivalent NaN-free fallback. Note: boundary values shift slightly compared to previous releases for users on the default settings. (#616)

• BoundaryForcing gains a unified prefill argument (with prefill_kwargs) to optionally fill the source before regridding: None (default, no prefill), "2d_lateral_fill" (legacy AMG Poisson fill), "inverse_dist" / "nearest_s2d" (xESMF source-on-source fills), or "nearest_neighbor" (cheap scipy distance-transform fill; also the automatic fallback when xESMF is unavailable). The horizontal regrid engine is now selected independently via regrid_method ("auto" (default; xESMF if installed else scipy), "xesmf", or "scipy"). (#616)

• New type is added to SurfaceForcing. restoring creates restoring forces files for ROMS (‘sss’ is the only option) (#589)

• CDRForcing has an option time_interpolation to choose step-like or interpolated releases (#601)

• RiverForcing supports time-varying river BGC from the RIVR2O export product via bgc_source={"name": "RIVR2O", "path": ...}; MARBL tracers not supplied dynamically are filled from bgc_source["fill"] (default: recommended constants from river_tracer_defaults.nc) (#615)

• RIVR2O concentrations vary by calendar year and are discharge-partitioned when rivers share a grid cell; missing annual files are linearly interpolated on the year axis (#615)

• CDR metrics report CO2 uptake as mass in tonnes of CO2 for both the air-sea flux-difference and DIC-difference constructions (using CO2 molar mass and native flux or inventory units when available). (#591)

• The CDR metrics figure from ROMSOutput.cdr_metrics() uses twin y-axes: CDR efficiency (dimensionless) and CO2 uptake (tonnes CO2), with title “CO2 uptake and CDR efficiency”. (#591)

• SurfaceForcing with type='restoring' now allows for making DIC and ALK restoring forces files (#619)

• SurfaceForcing gains the same regrid/prefill controls as BoundaryForcing (all types): regrid_method ("auto" (default; xESMF if installed else scipy), "xesmf", "scipy"), prefill (with prefill_kwargs; None default = no source fill, or "2d_lateral_fill"/"nearest_neighbor"/"inverse_dist"/"nearest_s2d"/"creep_fill"), and extrap_method (with extrap_kwargs) for the default no-prefill path. These are recorded in the saved dataset metadata and round-trip through to_yaml/from_yaml. (#XXX)

• SurfaceForcing with correct_radiation=True now scales to long (multi-year) records: the ERA5 radiation-correction climatology is interpolated onto the forcing time axis in chunks, so the correction no longer materializes the full (time, y, x) field at once. This also keeps per-timestep access (e.g. validation) cheap. (#629)

• New “Downloading ERA5 data” section in the datasets documentation shows how to pre-save ERA5 surface forcing (one file per day, NetCDF or Zarr) from the analysis-ready ARCO cloud store for offline/local use. (#629)

• Added support for the GloFAS v4.0 global river discharge dataset as an alternative to Dai & Trenberth, along with performance improvements and bug fixes to the river forcing pipeline. (#625)

• The end-to-end example notebook can now be downloaded as a plain Python script from the docs (#630)

Internal Changes

• A function to create a dict from a Grid obj (#573)

• A function to check if child wet points are outside the parent (#573)

• ChildGrid tests were adapted to the new framework (#573)

• Regression tests for ChildGrid (#550)

• Consolidate vertical regridding objects (#528)

• Enforce int32 type on ChildGrid mask (#559)

• Ensure regionmask>=0.11.0 in pyproject.toml (#565)

• Integrate narrow-channel closing directly into Grid.update_mask() (internal _close_narrow_channels), iterating north–south and east–west up to 10 passes

• short and long wave radiation time is shifted 1/2 a timestep sooner and have a dim of rad_time (#586)

• The coarse UNIFIED BGC dataset used for testing was updated to have depths of 0 and 5 m available (#589)

• 2 checks added for a point source when plotting CDRForcing.plot_distribution(). Low hsc is treated as a point source (#600)

• River BGC refactored behind a RiverBGCDataset protocol; RiverForcing merges dynamic and fill tracers via fill_river_bgc_concentrations(#615)

• Rivr2oRiverBGCDataset handles yearly file loading, spatial sampling, export-to-concentration conversion, and MARBL tracer mapping (#615)

• Monthly discharge climatology is expanded to calendar mid-month dates when combined with year-indexed RIVR2O BGC (expand_monthly_climatology_time_axis) (#615)

• CDR analysis tests cover tonnes-CO2 uptake and renamed efficiency fields; compute_cdr_metrics requires DIC_source in the input dataset alongside other CDR variables. (#591)

Documentation

• Nesting notebook is updated to match refactoring of ChildGrid to Grid objects. (#573)

• Move “overview of ROMS-Tools functionality” section from paper to docs (#554)

• Document close_narrow_channels option in Grid and update_mask(); update notebook examples

• The boundary forcing notebook is updated to describe the new NaN-aware horizontal regridding (masked bilinear with inverse-distance extrapolation), replacing the previous “1D versus 2D horizontal fill” discussion, and adds a section comparing the prefill source-fill methods (2d_lateral_fill, inverse_dist, nearest_s2d, nearest_neighbor) with a strategy table and side-by-side plots of the pre-filled source.

• Both the surface forcing and datasets notebooks are updated to reflect restoring function and WOA data (#589)

• The cdr notebook is updated to reflect interpolation option. Default is same as ROMS, no interpolation (#601)

• river_forcing.ipynb documents RIVR2O BGC forcing, including climatological discharge with year-varying tracers (#615)

• CDR analysis notebook: describe CO2 uptake (tonnes CO2) and CDR efficiency together, consistent with the metrics and figure. (#591)

• Notebooks updated to document density-space BGC interpolation (#620)

• Remove example support for CESM data. Boundary, Surface, Initial. (#633)

• Added subchunking example code and information on new datasets docs page. (#627)

Bugfixes

• Rotate ROMS velocities before regridding (#522)

• Corrected enclosed-basin filling in mask generation by iterating connected-component labels 1..nreg in _fill_enclosed_basins, preventing spurious interior lakes; updated the enclosed-region test to expect a single connected wet region.

• Corrected enclosed-basin filling in mask generation by iterating connected-component labels 1..nreg in _fill_enclosed_basins, preventing spurious interior lakes; updated the enclosed-region test to expect a single connected wet region. (#577)

• Fix timer logging messages during mask generation so durations render correctly when closing narrow channels and filling enclosed basins

• Fix for hanging when using the default of streaming from Copernicus for GLORYS output. (#604)

• Fix for river forcing indexing, sorting and plotting so that river_forcing.plot() and river_forcing.plot_locations() correctly shows top 20 largest rivers when plotting.

v3.5.0

New Features

• Initialize inner-nest initial conditions from outer-nest ROMS restart file (#513)

v3.4.0

New Features

• Default boundary selection respects land-only boundaries for BoundaryForcing and ChildGrid (#504)

• Partition eta_u and xi_v dimension if present (#526)

• Add support for EMOD bathymetry (#531)

Breaking Changes

• Remove user-facing control of SSH-adjusted depth coordinates (adjust_depth_for_sea_surface_height parameter) for InitialConditions and BoundaryForcing, enforcing consistent and physically sensible defaults (#514)

• Remove horizontal_chunk_size parameter from InitialConditions; the new vertical regridding scheme is sufficiently memory-efficient without it (#528)

Internal Changes

• Introduce ROMSDataset class analogous to LatLonDataset (#507)

• LateralFill: enforce correct dimension order for mask and input arrays (#517)

• Refactor rotate_velocities and staggered interpolation (#527)

Bugfixes

• Allow negative tracer flux values for CDRForcing (#510)

• Improve robustness of Grid.update_topography for externally generated grids (#523)

v3.3.0

New Features

• Plot wide grids with different cartopy projection (#496)

• Allow child domains with land points outside parent (#491)

• Add apply_mask option to plot function (#491)

• Improved bathymetry smoothing to better match UCLA MATLAB tools (#506)

Internal Changes

• Reorder operations in ChildGrid for correctness and clarity (#491)

• Update ChildGrid.plot_nesting to support nesting configurations where the child grid extends beyond the parent domain (#491)

• Rename Dataset to LatLonDataset and modularize RiverDataset (#500)

Documentation

• Document support for child grids that extend beyond parent grids (#491)

Bugfixes

• Make unit handling in RiverForcing.plot_locations() robust for grids without mask_rho units (#499)

• Only update child boundary indices if boundary not entirely on land (#503)

v3.2.0

New Features

• Enhance partition_netcdf flexibility: now supports partitioning multiple input files and allows specifying a custom output directory for saved tiles (#394)

• Add join_netcdf function: users can now re-join ROMS output and partitioned files (#412)

• Option to plot bathymetry along fixed lat/lon section (#399)

• Truly centered Gaussian CDR releases, reflecting recent changes in ucla-roms (#413)

• Introduce GLORYS streaming support for InitialConditions using copernicusmarine — not recommended yet; performance remains a blocker (#424)

• Allow performing CDR analysis, including plotting CDR uptake curves (#417)

• Introduce parameter allow_flex_time for InitialConditions controlling how strictly ini_time is handled (#460)

• Add utility function get_glorys_bounds to return the min/max latitude and longitude required for GLORYS downloads for a fixed domain (#445)

• Add compute_total_cdr_source method to CDRForcing, which computes integrated tracer quantities for all releases and returns a pandas DataFrame including a units row (#444)

• Plot ensembles of CDR uptake curves and their mean and standard deviation (#467)

• Allow custom shapefiles (such as GSHHS) to be used for coastline masking (#475)

Breaking Changes

• Default of new allow_flex_time parameter for InitialConditions is set to False. Old behaviour was allow_flex_time = True. (#460)

Documentation

• Add an end-to-end example to help users run through the full workflow (#468)

Bugfixes

• Fix bug in creation of InitialConditions (#460)

• Remove coordinate variables in nesting info file that conflict with ROMS expectations (#487)

• Allow regridding of 2D fields (#485)

v3.1.2

New Features

• Enhance partition_netcdf flexibility: now supports partitioning multiple input files and allows specifying a custom output directory for saved tiles (#394)

• Option to plot bathymetry along fixed lat/lon section (#399)

• Truly centered Gaussian CDR releases, reflecting recent changes in ucla-roms (#413)

Internal Changes

• Require xgcm version >= 0.9.0 (#425)

Documentation

• Fix broken link to GLORYS dataset (#397)

• Include recommendation about interp_frc in INFO message (#352)

• Document downloading of external datasets (#445)

v3.1.1

Internal Changes

• Add zarr as optional dependency (#393)

v3.1.0

New Features

• Allow plotting of a selected subset of river locations or time series, using the river_names argument (#389)

• Limit plotting to a maximum of 20 rivers/CDR forcings to match colormap constraints, with a warning if more are provided (#389)

• A unified plot function is now available, enabling users to create custom plots more easily (#375)

from roms_tools.plot import plot
plot(field, grid.ds)

• Gaussian CDR plots replicate the internal behavior of ROMS more accurately (#379)

• Section plots now include gray topography shading (#379)

• Grid latitude/longitude generation is updated to match UCLA MATLAB tools (#381)

• Bathymetry smoothing is modified by reducing the smoothing scale and applying an area-weighted scheme (#381)

• Remove cartopy coastlines from plots for the sake of clarity (#387)

Breaking Changes

• Grid.plot() no longer accepts the bathymetry argument. Bathymetry is now always plotted by default. (#375)

Internal Changes

• Most classes now delegate their .plot() methods to the centralized plot function, reducing code duplication and simplifying maintenance. (#375)

• Implement ruff rules (#382)

• Implement streamable default datasources for GLORYS (#386)

Documentation

Bugfixes

• Prevent plotting errors when more than 20 CDR or river forcings are present (#389)

• Fix bug incorrectly identifying CDR releases as outside the domain (#377)

• Add a de-duplication step that ensures that river names are unique (#378)

• Grid boundary plotting now provides a more accurate and consistent visual representation, with a default edge color of black. (#375)

• Handle overlapping rivers correctly (#356)

v3.0.0

New Features

• Visualize Gaussian distribution associated with CDR releases (#345)

• Make Grid.from_file() more robust for non-ROMS-Tools-generated grids (#365)

• Optimize wind drop-off calculation to avoid out-of-memory errors (#367)

• Option to stream ERA5 data directly from the cloud so that users do not have to pre-download ERA5 data (#357)

• Option for wind drop-off near the coasts in SurfaceForcing (#351)

• Option to ignore coarse dimensions when partitioning (#348)

• Ensure clean initialization of SurfaceForcing with ERA5 data (no warnings) (#337)

• Handle duplicate time entries in source data (#336)

• Set BGC tracers in rivers to non-zero default values (#326)

• Add .get_tracer_metadata() method to VolumeRelease and TracerPerturbation to allow users to inspect expected tracer units ahead of time (#327)

• Include tracer units in CDR forcing YAML files (#327)

• Add CDRForcing as a unified interface for CDR releases with and without volume flux (#301)

• Introduce TracerPerturbation for tracer-only CDR scenarios (#301)

• Add .plot_tracer_flux() method for visualizing tracer flux time series (#301)

• Allow bigger grid sizes of up to 25000 km (#311)

• Allow larger bottom control parameter theta_b up to 10 (#317)

Breaking Changes

• Remove class CDRVolumePointSource (#301)

• Require users to explicitly construct VolumeRelease and TracerPerturbation objects (#301)

• Drop support for Python 3.10 (#309)

Internal Changes

• Introduce Pydantic Release object refactoring the CDRVolumePointSource (#298)

• Add zero padding to partition_netcdf file numbers in filenames (#300)

• Refactor Release into its own module with subclassing (VolumeRelease, TracerPerturbation) (#301)

• Add new core classes: ReleaseSimulationManager, ReleaseCollector, and CDRForcingDatasetBuilder (#301)

Documentation

• Update example notebook demonstrating new CDRForcing workflow and release configuration (#301)

• Example notebooks now use updated unified BGC datasets (#320)

Bugfixes

• Fix plotting ROMSOutput for grids that straddle the dateline (#347)

• Report topography source path in NetCDF grid file so that the sequence Grid.from_file() –> grid.to_yaml() works (#353)

• Fix bug related to passing an ax to plotting methods (#325)

• Fix bug for grid = Grid.from_file() –> grid.to_yaml() sequence (#334)

• Fix handling of optional variables in Unified BGC datasets (#320)

v2.7.0

New Features

• New class CDRVolumePointSource for creating Carbon Dioxide Removal (CDR) forcing in ROMS simulations. It supports point-source injection of water and BGC tracers at fixed locations, designed for field-scale deployments, where localized mixing is essential (#295)

• TidalForcing class now works with original (rather than postprocessed) TPXO data (#254)

Breaking Changes

• TidalForcing class now expects original (rather than postprocessed) TPXO data. See documentation for details. (#254)

Internal Changes

• Correct default value for NOx from 1e-13 to 1e-12 kg/m2/s (#294)

• The TidalForcing class now correctly handles staggered grid of TPXO data (#254)

• The TidalForcing class now includes enhanced and more robust constituent correction, making it compatible with newer TPXO products (#254)

• The Self-Attraction and Loading (SAL) correction for the tidal forcing is sourced internally from the TPXO09v2a dataset since it is not included in the regularly updated TPXO datasets (#254)

Documentation

• Updated documentation for the TidalForcing class and its dataset requirements (#254)

v2.6.2

New Features

• Enable reading from unified BGC dataset (#274)

Internal Changes

• Refactoring of InitialConditions, BoundaryForcing, and SurfaceForcing to accommodate optional variable names (#274)

• Modification of the Dataset class including the choose_subdomain method, the capability to handle fractional days in a climatology, and the addition of a needs_lateral_fill attribute (#274)

• Separation of river_flux variable into river_index and river_fraction (#291)

v2.6.1

New Features

• Support to regrid ROMS output data onto lat-lon-z grid (#286)

Internal Changes

• Rename river_location to river_flux (#283)

Bugfixes

v2.6.0

New Features

• Support to plot ROMS output data at lat/lon locations (#277)

• Support to plot ROMS output data along sections of fixed latitude or longitude (#278)

• Support to plot ROMS output data at fixed depth (#279)

• Support for saving a figure of ROMS output data (#280)

Internal Changes

• Unfreeze arguments in all dataclasses (#276)

• Integration with xesmf for horizontal regridding from ROMS (#277)

• Computation of nominal horizontal resolution in degrees (#278)

• Integration with xgcm and numba for vertical regridding from ROMS (#279)

v2.5.0

New Features

• Support for creating multi-cell rivers (#258)

• Support for writing and reading single-cell and multi-cell rivers to/from YAML (#258)

• Enable plotting ROMS output without boundary; helpful because boundary for ROMS diagnostics consists of zeros (#265)

• Nicer y-labels for depth plots (#265)

• Option to enable or disable adjusting for SSH in depth coordinate calculation for ROMSOutput (#269)

Breaking Changes

• Deprecate type parameter in ROMSOutput (#253)

• Write and read the parameter bypass_validation to/from YAML (#249)

• Refactor Nesting class and renamed it to ChildGrid class to ensure definite serialization (#250)

Internal Changes

• Enforce double precision on source data to ensure reproducible results (#244)

• Results produced with vs. without Dask in test suite now pass with xr.testing.assert_equal confirming reproducibility (#244)

• Document the option for start_time = None and end_time = None in the docstrings for BoundaryForcing and SurfaceForcing, specifying that when both are None, no time filtering is applied to the data. Also, ensure a warning is raised in this case to inform the user. (#249)

• Move conversion to double precision to after choosing subdomain of source data, ensuring a speed-up in grid generation and other forcing datasets that do not use Dask (#264)

Documentation

• Documentation on how to use ROMS-Tools with Dask (#245)

• More detailed documentation of ROMSOutput (#269)

Bugfixes

v2.4.0

New Features

• Introduce new parameter coarse_grid_mode for SurfaceForcing. The default coarse_grid_mode = "auto" automatically decides whether it makes sense to interpolate onto the coarse grid, which saves computations (#228)

• New default for correct_radiation in SurfaceForcing is True (#228)

• New default for bathymetry in Grid.plot() is True (#234)

• New default for group in SurfaceForcing.save() and BoundaryForcing.save() is True (#236)

• Option to adjust depth for sea surface height when creating InitialConditions and BoundaryForcing (#240)

• New parameter horizontal_chunk_size for InitialConditions, which ensures the feasibility of processing initial conditions for large domains, both in terms of memory footprint and compute times (#241)

Breaking Changes

• Remove support for partitioning files upon saving (#221)

• Remove parameter use_coarse_grid for SurfaceForcing class (#228)

• Remove parameter filepath_grid from RiverForcing.save() method (#232)

Internal Changes

• Parallelize computation of radiation correction, leading to a hugely improved memory footprint for surface forcing generation (#227)

• For computation of radiation correction, swap order of temporal and spatial interpolation to further improve memory footprint (#227)

• Write river locations into river forcing file rather than into grid file (#232)

• When appropriate, only compare data hashes rather than file hashes in automated test suite (#235)

• Slightly shift one of the test grids away from the Greenwich meridian (#235)

• The partitioning functions are moved to their own subdirectory tiling (#236)

• Internal refactoring of depth coordinate computation (#240)

Documentation

• New features and defaults are documented for SurfaceForcing ((#228)

• Improvements to the notebook that documents the partioning functionality (#236)

• Document the option to adjust depth for sea surface height (#240)

• More realistic, higher-resolution domains in the example notebooks, which use the SRTM15 topography to discourage the user from employing the default ETOPO5 topography (#241)

Bugfixes

• Fix bug in validation step for surface forcing (#227)

v2.3.0

New Features

• ROMSOutput class for analyzing ROMS model output (#217)

Bugfixes

• Correctly handle list of files in source and bgc_source in YAML files (#218)

v2.2.1

Bugfixes

• Correctly write Pathlib bgc_source to YAML file (#215)

v2.2.0

Bugfixes

• Fix bug in validation of tidal forcing which led to memory blow-ups (#211)

• Make sure correct masks are used for validation (#214)

• Correctly write Pathlib objects to YAML file (#212)

v2.1.0

New Features

• Nesting capability (#204)

• Option to bypass validation (#206)

• Latitude and longitude labels are added to spatial plots (#208)

Internal Changes

• Dirty tagging in the versioning has been removed (#202).

Documentation

• Release notes have been added to the documentation (#202).

Bugfixes

• Bug with radiation correction in surface forcing has been fixed (#209).

v2.0.0

New Features

• This release supports more accurate topography regridded from SRTM15, but also still supports the previous ETOPO5 topography (#172).

• The land mask is improved since it is now inferred from a separate coastline dataset rather than from the regridded topography (#172).

Breaking Changes

• The topography format is no longer backward compatible. The syntax for specifying the topography source has changed (#172). The new format is

grid = Grid(topography_source={"name": "ETOPO5"})

instead of the previous format

grid = Grid(topography_source="ETOPO5")

Internal Changes

• Topography regridding function has been generalized (#172).

• The classes roms_tools.setup.datasets.ETOPO5Dataset and roms_tools.setup.datasets.SRTM15Dataset have been introduced (#172).

• The roms_tools.setup.datasets.Dataset class has been made more performant by

  • ensuring that global datasets are only concatenated along the longitude dimension when necessary and

  • restricting to the necessary latitute range first before concatenation (#172).

• Layer and interface depths are not computed as part of grid generation any longer because this can blow memory for large domains. Instead, layer and interface depths are only computed when needed as part of initial conditions and boundary forcing generation (#172).

Documentation

• SRTM15 topography examples have been added to the documentation (#172).

Bugfixes